Multi-agent evolution improves fake news detection with large language models
Reasoning through Evolution: Automatic Meta-path Discovery for LLM-based Fake News Detection
Artificial IntelligenceMachine Learning
Summary
Detecting fake news is hard because news spreading patterns are complex and current methods need lots of labeled examples. The researchers found a way to summarize these complex spreading patterns into simpler parts called meta-paths that help large language models (LLMs) understand the structure better. They use a method inspired by genetic evolution to find the best meta-paths automatically. This helps LLMs make better decisions about whether news is fake, even when there are few labeled examples available.
What this means in practice
- •For social media platform engineers: Enhance fake news detection modules by enabling LLMs to reason on simplified propagation patterns without large labeled datasets.
- •For cybersecurity analysts: Use evolved meta-paths for more efficient detection of misinformation spreading patterns in threat intelligence.
Authors
Ziyi Zhou, Xiaoming Zhang, Hui Pang, Yuting Zhang, Tiesunlong Shen, Bingyu Yan, Erik Cambria, Litian Zhang
Abstract
Propagation structures provide crucial evidence for fake news detection, yet existing approaches primarily rely on supervised GNN-based models, which require substantial labeled data and exhibit limited generalization. Although large language models (LLMs) exhibit strong reasoning capabilities, directly feeding them raw propagation graphs creates a significant modality mismatch and severe information overload, making structure-aware reasoning unreliable in zero-shot and few-shot settings. To bridge this gap, we propose MAGER, a multi-agent genetic evolution framework that automatically discovers meta-paths optimized for LLM reasoning. By compressing complex propagation graphs into informative subgraphs, the evolved meta-paths alleviate both information overload and modality mismatch, enabling frozen LLMs to perform structure-aware veracity reasoning. We further introduce a graph in-context learning strategy that retrieves semantically and structurally similar demonstrations to strengthen classification and reasoning. Extensive experiments show that MAGER substantially improves frozen LLMs as standalone fake news detectors in data-efficient settings. Our code is available at https://github.com/SenticNet/MAGER.